Wind tunnel operation auxiliary decision-making method, system and device based on pneumatic data monitoring

Abstract

The present invention discloses a method, system and device for auxiliary decision-making of wind tunnel operation based on pneumatic data monitoring. The method includes: acquiring historical pneumatic data of the wind tunnel and supplementing abnormal pneumatic data; defining a pneumatic data structure applicable to artificial intelligence methods, and performing feature engineering processing on normal pneumatic data and abnormal pneumatic data to obtain processed data; constructing multiple binary classification mathematical models for pneumatic data using multiple methods to judge whether the pneumatic data is normal or abnormal; automatically classifying and judging the pneumatic data generated in real time during the wind tunnel test; when all the multiple binary classification mathematical models for pneumatic data judge that the pneumatic data generated in real time by the wind tunnel is normal, it is determined that the pneumatic data generated in this wind tunnel test is normal, and the wind tunnel test of the next state is carried out; otherwise, the wind tunnel staff is reminded to check for problems and the decision is made manually. The present invention can realize auxiliary decision-making of wind tunnel operation based on pneumatic data monitoring.

Description

Technical Field

[0001] The present invention relates to the field of wind tunnel operation auxiliary decision-making, and particularly to a wind tunnel operation auxiliary decision-making method, system and device based on pneumatic data monitoring. Background Art

[0002] A wind tunnel is a large-scale ground simulation device for aerodynamics. According to the similarity principle, a scaled-down aircraft model or physical object is placed in the wind tunnel. Based on the principle of relativity, the air flow velocity is artificially controlled to simulate the relative motion between the aircraft and the air flow, and various complex flight states of the aircraft are simulated in the wind tunnel to obtain test data corresponding to the states.

[0003] The attitude of a real aircraft in the air changes continuously, and the flight speed of the aircraft also changes continuously in real time. In wind tunnel tests, the single-factor method is mostly used, that is, during a wind tunnel test, only one degree of freedom of the aircraft changes, and other degrees of freedom and air flow velocity adopt fixed values. By conducting multiple wind tunnel tests and selecting different fixed values in each wind tunnel test to construct the entire discrete state space, and finally obtaining continuous pneumatic data for the entire space through interpolation.

[0004] Wind tunnel tests are a complex system integrating multiple disciplines. The complexity of wind tunnel test implementation and the extremely harsh test environment make it possible for wind tunnel test data to be abnormal or even incorrect. Whether the wind tunnel test data is true and reliable is directly related to the success or failure of model development and flight tests. Coupled with the fact that wind tunnel tests themselves are high-energy-consuming and high-cost tests, it is crucial to accurately judge the pneumatic test data generated in each wind tunnel test in a timely manner. In the traditional wind tunnel operation mode, the pneumatic test data generated in each wind tunnel operation is diagnosed manually by personnel, which requires the personnel to have extremely rich expert experience and to always maintain a high degree of concentration. However, because a wind tunnel is a scarce resource and usually operates continuously 7×24 hours throughout the year, the personnel are also in a high-load operation process all the time, and it is difficult to avoid misjudgment.

[0005] By statistically analyzing the wind tunnel test data over the years, it can be concluded that the proportion of normal pneumatic data is greater than 95%, and the proportion of abnormal pneumatic data is very small. Because once abnormal data with a very small proportion appears, if it cannot be discovered in time, it will cause huge losses, resulting in the current wind tunnel test tasks investing a huge amount of manpower and material resources to discover small probability events. Summary of the Invention

[0006] The purpose of the present invention is to provide a wind tunnel operation auxiliary decision-making method, system and device based on pneumatic data monitoring, aiming to solve the problem of wind tunnel operation auxiliary decision-making.

[0007] The present invention provides a method for assisting in decision-making for wind tunnel operation based on pneumatic data monitoring, including:

[0008] S1. Obtain the historical pneumatic data of the wind tunnel, and according to the labels of normal or abnormal data, obtain the normal pneumatic data and abnormal pneumatic data, and construct a normal pneumatic database and an abnormal pneumatic database based on the normal pneumatic data and abnormal pneumatic data;

[0009] S2. Supplement the abnormal pneumatic data according to the typical causes of wind tunnel abnormal data and the method for generating abnormal data;

[0010] S3. Define a pneumatic data structure suitable for artificial intelligence methods, and perform feature engineering processing on the normal pneumatic data and abnormal pneumatic data to obtain processed data;

[0011] S4. Use multiple methods to construct multiple binary classification mathematical models for judging whether the pneumatic data is normal or abnormal based on the processed data;

[0012] S5. Embed the multiple binary classification mathematical models for pneumatic data into the wind tunnel operation system to automatically classify and judge the pneumatic data generated by the wind tunnel test in real time;

[0013] S6. When all the multiple binary classification mathematical models for pneumatic data judge that the pneumatic data generated by the wind tunnel in real time is normal, it is determined that the pneumatic data generated by this wind tunnel test is normal, and the wind tunnel test in the next state is carried out; otherwise, it is determined that there may be an abnormality in this wind tunnel test, and the wind tunnel staff is reminded to check the problem, and the decision is made manually.

[0014] The present invention also provides a wind tunnel operation assistance decision-making system based on pneumatic data monitoring, including:

[0015] An acquisition module, configured to acquire the historical pneumatic data of the wind tunnel, and according to the labels of normal or abnormal data, obtain the normal pneumatic data and abnormal pneumatic data, and construct a normal pneumatic database and an abnormal pneumatic database based on the normal pneumatic data and abnormal pneumatic data;

[0016] A supplement module, configured to supplement the abnormal pneumatic data according to the typical causes of wind tunnel abnormal data and the method for generating abnormal data;

[0017] A definition module, configured to define a pneumatic data structure suitable for artificial intelligence methods, and perform feature engineering processing on the normal pneumatic data and abnormal pneumatic data to obtain processed data;

[0018] A judgment module, configured to use multiple methods to construct multiple binary classification mathematical models for judging whether the pneumatic data is normal or abnormal based on the processed data;

[0019] An embedding module, which is used to embed a binary classification mathematical model of various aerodynamic data into a wind tunnel operation system to automatically classify and judge the aerodynamic data generated by wind tunnel tests in real time;

[0020] A real-time judgment module, which is used to determine that the aerodynamic data generated by this wind tunnel test is normal when the binary classification mathematical models of various aerodynamic data all judge that the aerodynamic data generated by the wind tunnel in real time is normal, and proceed with the wind tunnel test in the next state; otherwise, it is determined that there may be an abnormality in this wind tunnel test, and the wind tunnel staff is reminded to troubleshoot the problem, and the decision is made manually.

[0021] An embodiment of the present invention also provides a wind tunnel operation auxiliary decision-making device based on aerodynamic data monitoring, including: a memory, a processor, and a computer program stored on the memory and executable on the processor. When the computer program is executed by the processor, the steps of the above method are implemented.

[0022] An embodiment of the present invention also provides a computer-readable storage medium, on which an implementation program for information transmission is stored. When the program is executed by a processor, the steps of the above method are implemented.

[0023] By adopting the embodiment of the present invention, manpower is saved through a mathematical model, and the production efficiency of the wind tunnel is improved.

[0024] The above description is only an overview of the technical solution of the present invention. In order to be able to understand the technical means of the present invention more clearly, it is implemented in accordance with the content of the description. And in order to make the above and other purposes, features and advantages of the present invention more obvious and understandable, the following specifically describes the embodiments of the present invention. Description of the Drawings

[0025] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. Obviously, the following drawings are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0026] Figure 1 is a flowchart of the wind tunnel operation auxiliary decision-making method based on aerodynamic data monitoring according to an embodiment of the present invention;

[0027] Figure 2 is a specific flowchart of the wind tunnel operation auxiliary decision-making method based on aerodynamic data monitoring according to an embodiment of the present invention;

[0028] Figure 3 is a schematic diagram of the wind tunnel operation auxiliary decision-making system based on aerodynamic data monitoring according to an embodiment of the present invention;

[0029] Figure 4 It is a schematic diagram of an auxiliary decision-making device for wind tunnel operation based on pneumatic data monitoring according to an embodiment of the present invention. Specific Embodiments

[0030] Next, the technical solutions of the present invention will be clearly and completely described in conjunction with the embodiments. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0031] Method Embodiment

[0032] According to an embodiment of the present invention, there is provided an auxiliary decision-making method for wind tunnel operation based on pneumatic data monitoring. Figure 1 It is a flowchart of an auxiliary decision-making method for wind tunnel operation based on pneumatic data monitoring according to an embodiment of the present invention, as Figure 1 shown, and specifically includes:

[0033] S1. Obtain the historical pneumatic data of the wind tunnel, and according to the labels of normal or abnormal data, obtain the normal pneumatic data and abnormal pneumatic data, and construct a normal pneumatic database and an abnormal pneumatic database based on the normal pneumatic data and abnormal pneumatic data;

[0034] S2. Supplement the abnormal pneumatic data according to the typical causes of wind tunnel abnormal data and the method of generating abnormal data;

[0035] S3. Define a pneumatic data structure applicable to artificial intelligence methods, and perform feature engineering processing on the normal pneumatic data and abnormal pneumatic data to obtain processed data;

[0036] S3 specifically includes: defining a pneumatic data structure applicable to artificial intelligence methods, and the data structure includes four categories of data: wind tunnel test status, basic model information, aerodynamic characteristics, and algorithm label extension interface, and performing feature engineering processing on the four categories of data to obtain processed data.

[0037] S4. Use multiple methods to construct multiple binary classification mathematical models for judging whether the pneumatic data is normal or abnormal based on the processed data;

[0038] S4 specifically includes:

[0039] Construct a binary classification mathematical model for judging whether the pneumatic data is normal or abnormal using the isolation forest method, select the normal pneumatic data and abnormal pneumatic data as training samples, and ensure that the proportion of abnormal data does not exceed 20%, and mix the normal and abnormal data disorderly together, and do not use the normal or abnormal data labels during the training process;

[0040] A binary classification mathematical model for judging whether pneumatic data is normal or abnormal is constructed using a single-class support vector machine, and only normal pneumatic data is selected as the training sample;

[0041] A binary classification mathematical model for judging whether pneumatic data is normal or abnormal is constructed using a deep neural network. Normal pneumatic data and abnormal pneumatic data are selected as the training samples, and normal or abnormal data labels are used during the training process.

[0042] S5. Embed multiple binary classification mathematical models of pneumatic data into the wind tunnel operation system to automatically classify and judge the pneumatic data generated by wind tunnel tests in real time;

[0043] S6. When multiple binary classification mathematical models of pneumatic data all judge that the pneumatic data generated by the wind tunnel in real time is normal, it is determined that the pneumatic data generated by this wind tunnel test is normal, and the wind tunnel test of the next state is carried out; otherwise, it is determined that there may be abnormalities in this wind tunnel test, and the wind tunnel staff is reminded to check the problems, and the decision is made manually.

[0044] Figure 2 It is a specific flowchart of the wind tunnel operation auxiliary decision-making method based on pneumatic data monitoring in the embodiment of the present invention;

[0045] The purpose of the present invention is to provide an intelligent auxiliary decision-making method for wind tunnel operation based on pneumatic data monitoring. This method can make wind tunnel operation decisions efficiently and with high quality, reduce the workload of wind tunnel staff, reduce the misjudgment rate, save the cost of repeated wind tunnel tests, and improve the efficiency of wind tunnel tests.

[0046] An intelligent auxiliary decision-making method for wind tunnel operation based on pneumatic data monitoring specifically includes the following steps:

[0047] (1). Standardize and organize historical pneumatic data, and construct a normal pneumatic database and an abnormal pneumatic database according to the labels of normal or abnormal data;

[0048] (2). To address the problem that the scarcity of abnormal pneumatic data cannot meet the requirements of artificial intelligence methods, artificial abnormal pneumatic data is supplemented based on the typical causes of wind tunnel abnormal data and abnormal data generation methods.

[0049] The typical causes of wind tunnel abnormal data include: balance failure, acquisition system failure, incorrect model installation, incorrect flow field control, etc.; based on the causes of abnormal data generation and analyzing its mechanism, abnormal data generation methods can be constructed. The most commonly used abnormal data generation method is to correct normal data, including: adding an offset to normal data, fluctuating one or several points in the normal data sequence, and the data from a certain point onwards in the normal data sequence no longer changing, etc.;

[0050] After the above two steps, the number of finally constructed databases is shown in Table 1:

[0051] Table 1

[0052] Category Total number Training sample Test sample Normal 100000 80000 20000 Abnormal 10000 8000 2000

[0053] (3) Define the aerodynamic data structure applicable to the artificial intelligence method, and perform feature engineering processing on the data;

[0054] The aerodynamic data structure applicable to the artificial intelligence method is characterized in that it includes four categories of data: wind tunnel test status, basic model information, aerodynamic characteristics, and label extension interface. The specific data format is shown in Table 2:

[0055] Table 2

[0056] Serial number 1 2 3 4 5 6 7 8 Symbol CC <![CDATA[M a > <![CDATA[P a > <![CDATA[P0]]> <![CDATA[P ∞ > T <![CDATA[R e > <![CDATA[α corr > Physics Train number Mach Atmosphere Total pressure Static pressure Total temperature Reynolds Airflow Serial number 9 10 11 12 13 14 15 16 Symbol V <![CDATA[G0]]> Sym L D Face <![CDATA[D p > Roll Physics Model Center of mass Axial symmetry Projectile length Diameter Facing Rudder deflection Rolling Serial number 17 18 19 20 21 22 23 24 Symbol Beta My Z Mz Y Mx Q α Physics Sideslip Yaw Lateral direction Pitch Pitch Roll Axial direction Angle of attack Serial number 25 26 27 28 Symbol Label IF SVM DNN Physics Artificial Algorithm Algorithm Algorithm

[0057] In Table 2, the numerical values of the parameters corresponding to serial numbers 1-24 can all be obtained by sorting out the wind tunnel historical data; the Label of serial number 25 is the data label manually marked. When the aerodynamic data is normal, the Label value is 0. When the aerodynamic data is abnormal, the Label value is 1. When the manual annotation is not performed, the Label value is 99; the IF, SVM, and DNN of serial numbers 26-28 are the data labels for annotating the data by the Isolation Forest, One-Class Support Vector Machine, and Deep Neural Network methods respectively. When the aerodynamic data is normal, the set value is 0. When the aerodynamic data is abnormal, the set value is 1. When the method does not perform annotation, the set value is 99; the parameters corresponding to serial numbers 18-24 are vectors, and the vector length is equal to the angle of attack sequence length of a single wind tunnel test, and the vector length is not fixed; the parameters corresponding to serial numbers 1-17 and 25-28 are scalars;

[0058] (4) Select three methods of Isolation Forest, One-Class Support Vector Machine, and Deep Neural Network to construct binary classification mathematical models for aerodynamic data respectively. The three constructed binary classification mathematical models for aerodynamic data can respectively realize the judgment of whether the aerodynamic data is normal or abnormal;

[0059] The classification situation of the binary classification mathematical model is shown in Table 3:

[0060] Table 3

[0061]

[0062] It includes four situations: the first: actually normal and predicted as normal; the second: actually normal and predicted as abnormal; the third: actually abnormal and judged as normal; the fourth: actually abnormal and judged as abnormal. The most ideal classification effect is that the recognition rates of the first and fourth situations are both 100%, and the second and third situations are 0%;

[0063] The three methods of Isolation Forest, One-Class Support Vector Machine, and Deep Neural Network are selected to construct a binary classification mathematical model for aerodynamic data, characterized in that:

[0064] For the Isolation Forest method, normal aerodynamic data and abnormal aerodynamic data are selected as training samples, and it is ensured that the proportion of abnormal data does not exceed 20%. The normal and abnormal data are mixed randomly without using the data labels of normal or abnormal data during the training process. Selecting the proportion of abnormal data not to exceed 20% is based on the characteristics of the Isolation Forest method itself and the characteristics of aerodynamic data, and is a conclusion drawn based on expert experience and actual data testing.

[0065] For the One-Class Support Vector Machine, only normal aerodynamic data is selected as the training sample.

[0066] For the Deep Neural Network, normal aerodynamic data and abnormal aerodynamic data are selected as training samples, and the data labels of normal or abnormal data are used during the training process.

[0067] From the perspective of applying the training samples, the process of constructing the binary classification mathematical model by the above three methods can be considered independent. The training samples of the three methods are randomly selected from the training sample library, and the quantity distribution is shown in Table 4:

[0068] Table 4

[0069]

[0070] Because it is necessary to ensure that abnormal data cannot be judged as normal data, that is, the FP situation of the binary classification model needs to be 0%, and the FN situation needs to be 100%. Therefore, during the process of training the model, the situation of being judged as normal is very strict, and the situation of being judged as abnormal is relatively loose. The method parameters leave sufficient margins. The operation effects of the binary classification mathematical models obtained by the three methods on the training sets allocated in Table 4 on the test set are shown in Table 5:

[0071] Table 5

[0072]

[0073] (5) Embed the above three binary classification mathematical models for aerodynamic data into the wind tunnel operation system to realize real-time and automatic classification judgment of the aerodynamic data generated by wind tunnel tests;

[0074] (6) When the above three binary classification mathematical models for aerodynamic data all judge that the aerodynamic data generated in real time by the wind tunnel is normal, it is determined that the aerodynamic data generated by this wind tunnel test is normal, and the wind tunnel test of the next state can be carried out; otherwise, it is determined that there may be abnormalities in this wind tunnel test, reminding the wind tunnel staff to check the problems, and the final decision is made manually.

[0075] When all three binary classification models determine that the data is normal, the data is determined to be normal; otherwise, the data is considered abnormal. The statistical performance on the test set is shown in Table 6 below:

[0076] Table 6

[0077]

[0078] As can be seen from Table 6, only 71.06% of the originally normal data is determined to be normal, but the reliability of the method is guaranteed to a great extent.

[0079] By statistically analyzing the wind tunnel test data over the years, it can be concluded that the proportion of normal aerodynamic data is greater than 95%. According to the training samples and test samples given in this embodiment, 95% * 71.06% = 67.507% of the data proposed by the method of the present invention does not require manual participation in diagnosis, that is, the analysis efficiency of wind tunnel test data is increased by 67.507%. When different training samples, test samples, and the number of samples change, the above values will also change. However, based on the generally recognized experience of those skilled in the art, it can be concluded that the present invention can provide an auxiliary decision-making function for wind tunnel operation, reduce the work intensity of wind tunnel operators, reduce the misjudgment rate caused by manual judgment, and improve the efficiency and quality of wind tunnel operation.

[0080] The advantages of the present invention compared with the prior art are as follows:

[0081] The aerodynamic data structure applicable to the artificial intelligence method proposed by the present invention is simple and precise. It not only covers the necessary information required for the analysis of aircraft aerodynamic data, but also provides a convenient interface for the application of artificial intelligence methods in the aerodynamic field;

[0082] The present invention takes a different approach and does not pursue the performance index of the abnormal data detection rate. Instead, it aims to ensure that the data determined to be normal by the algorithm does not contain abnormal data. Starting from this point, it is convenient to be applied to the actual production of wind tunnels and has strong engineering practicability;

[0083] The present invention selects three simple and highly mature methods, namely Isolation Forest, One-Class Support Vector Machine, and Deep Neural Network, to respectively implement the binary classification mathematical model of aerodynamic data. The method is not complex and can be easily implemented by professionals in related fields. Moreover, all three methods can achieve the index that the data determined to be normal by the algorithm does not contain abnormal data through simple parameter tuning, making it easier to popularize and apply;

[0084] The reason why the present invention selects the three methods of Isolation Forest, One-Class Support Vector Machine, and Deep Neural Network is based on the idea that the probability of three mutually independent events occurring simultaneously in mathematics is relatively low. Although it reduces the probability of normal data being judged as normal, it greatly ensures the design criterion that the data judged as normal by the algorithm does not contain abnormal data, greatly ensures the correctness of the method for assisting decision-making, and enhances the confidence in engineering applications.

[0085] The present invention can play a role in assisting decision-making for wind tunnel operation, reducing the work intensity of wind tunnel operation personnel, reducing the misjudgment rate caused by manual judgment, and improving the efficiency and quality of wind tunnel operation.

[0086] System Embodiment

[0087] According to an embodiment of the present invention, there is provided a wind tunnel operation auxiliary decision-making system based on pneumatic data monitoring. Figure 3 It is a schematic diagram of the wind tunnel operation auxiliary decision-making system based on pneumatic data monitoring according to the embodiment of the present invention, as Figure 3 shown, specifically including:

[0088] An acquisition module, configured to acquire historical pneumatic data of the wind tunnel, obtain normal pneumatic data and abnormal pneumatic data according to the labels of normal or abnormal data, and construct a normal pneumatic database and an abnormal pneumatic database based on the normal pneumatic data and the abnormal pneumatic data;

[0089] A supplement module, configured to supplement abnormal pneumatic data according to the typical causes of wind tunnel abnormal data and the abnormal data generation method;

[0090] A definition module, configured to define a pneumatic data structure applicable to artificial intelligence methods, and perform feature engineering processing on the normal pneumatic data and the abnormal pneumatic data to obtain processed data;

[0091] A judgment module, configured to construct multiple binary classification mathematical models for judging the normality or abnormality of pneumatic data according to the processed data using multiple methods;

[0092] An embedding module, configured to embed multiple binary classification mathematical models of pneumatic data into the wind tunnel operation system to perform real-time and automatic classification judgment on the pneumatic data generated by the wind tunnel test;

[0093] A real-time judgment module, configured to, when multiple binary classification mathematical models of pneumatic data all judge that the pneumatic data generated by the wind tunnel in real time is normal, determine that the pneumatic data generated by this wind tunnel test is normal and proceed to the next state of the wind tunnel test; otherwise, determine that there may be an abnormality in this wind tunnel test, remind the wind tunnel staff to check the problem, and make a decision manually.

[0094] The definition module is specifically used for: defining a pneumatic data structure applicable to artificial intelligence methods, the data structure including four categories of data: wind tunnel test status, basic model information, aerodynamic characteristics, and algorithm label extension interfaces, and performing feature engineering processing on the four categories of data to obtain processed data.

[0095] The judgment module is specifically used for:

[0096] Constructing a binary classification mathematical model for judging the normality or abnormality of pneumatic data by using the Isolation Forest method, selecting normal pneumatic data and abnormal pneumatic data as training samples, and ensuring that the proportion of abnormal data does not exceed 20%, and mixing the normal and abnormal data disorderly together, without using the data labels of normal or abnormal data during the training process;

[0097] Constructing a binary classification mathematical model for judging the normality or abnormality of pneumatic data by using a one-class support vector machine, and only selecting normal pneumatic data as training samples;

[0098] Constructing a binary classification mathematical model for judging the normality or abnormality of pneumatic data by using a deep neural network, selecting normal pneumatic data and abnormal pneumatic data as training samples, and using the data labels of normal or abnormal data during the training process.

[0099] The embodiment of the present invention is a system embodiment corresponding to the above method embodiment. The specific operations of each module can be understood with reference to the description of the method embodiment, and will not be elaborated here.

[0100] Device Embodiment 1

[0101] The embodiment of the present invention provides a wind tunnel operation auxiliary decision-making device based on pneumatic data monitoring, as Figure 4 shown, including: a memory 40, a processor 42, and a computer program stored on the memory 40 and executable on the processor 42. When the computer program is executed by the processor, the steps in the above method embodiment are implemented.

[0102] Device Embodiment 2

[0103] The embodiment of the present invention provides a computer-readable storage medium, on which an implementation program for information transmission is stored. When the program is executed by the processor 42, the steps in the above method embodiment are implemented.

[0104] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not make the essence of the technical solutions of the embodiments of the present invention deviate from the scope of the present solution.

Claims

1. A wind tunnel operation auxiliary decision-making method based on pneumatic data monitoring, characterized in that including S1. Obtain the historical aerodynamic data of the wind tunnel. According to the labels of normal or abnormal data, obtain the normal aerodynamic data and abnormal aerodynamic data, and construct a normal aerodynamic database and an abnormal aerodynamic database based on the normal aerodynamic data and abnormal aerodynamic data; S2. Supplement the abnormal aerodynamic data according to the typical causes of wind tunnel abnormal data and the abnormal data generation method; S3. Define the aerodynamic data structure applicable to the artificial intelligence method, and perform feature engineering processing on the normal aerodynamic data and abnormal aerodynamic data to obtain the processed data. Specifically, it includes: Define the aerodynamic data structure applicable to the artificial intelligence method. The data structure includes four categories of data: wind tunnel test status, basic model information, aerodynamic characteristics, and algorithm label extension interface. Perform feature engineering processing on the four categories of data to obtain the processed data; S4. Use multiple methods to construct multiple binary classification mathematical models for judging whether the aerodynamic data is normal or abnormal based on the processed data. Specifically, it includes: Construct a binary classification mathematical model for judging whether the aerodynamic data is normal or abnormal using the Isolation Forest method. Select the normal aerodynamic data and abnormal aerodynamic data as training samples, and ensure that the proportion of abnormal data does not exceed 20%. Mix the normal and abnormal data randomly, and do not use the normal or abnormal data labels during the training process; Construct a binary classification mathematical model for judging whether the aerodynamic data is normal or abnormal using the One-Class Support Vector Machine. Only select the normal aerodynamic data as the training sample; Construct a binary classification mathematical model for judging whether the aerodynamic data is normal or abnormal using a deep neural network. Select the normal aerodynamic data and abnormal aerodynamic data as training samples, and use the normal or abnormal data labels during the training process; S5. Embed multiple binary classification mathematical models of aerodynamic data into the wind tunnel operation system to automatically classify and judge the aerodynamic data generated by the wind tunnel test in real time; S6. When multiple binary classification mathematical models of aerodynamic data all judge that the aerodynamic data generated by the wind tunnel in real time is normal, it is determined that the aerodynamic data generated by this wind tunnel test is normal, and the wind tunnel test of the next state is carried out; otherwise, it is determined that there may be an abnormality in this wind tunnel test, and the wind tunnel staff is reminded to check the problem, and the decision is made manually.

2. An auxiliary decision-making system for wind tunnel operation based on pneumatic data monitoring, characterized in that, including An acquisition module, which is used to obtain the historical aerodynamic data of the wind tunnel. According to the labels of normal or abnormal data, obtain the normal aerodynamic data and abnormal aerodynamic data, and construct a normal aerodynamic database and an abnormal aerodynamic database based on the normal aerodynamic data and abnormal aerodynamic data; A supplement module, which is used to supplement the abnormal aerodynamic data according to the typical causes of wind tunnel abnormal data and the abnormal data generation method; A definition module, which is used to define the aerodynamic data structure applicable to the artificial intelligence method, and perform feature engineering processing on the normal aerodynamic data and abnormal aerodynamic data to obtain the processed data. Specifically, it is used for: Define the aerodynamic data structure applicable to the artificial intelligence method. The data structure includes four categories of data: wind tunnel test status, basic model information, aerodynamic characteristics, and algorithm label extension interface. Perform feature engineering processing on the four categories of data to obtain the processed data; A judgment module, configured to use multiple methods to construct multiple binary classification mathematical models for judging whether pneumatic data is normal or abnormal based on the processed data; An embedding module, configured to embed the multiple binary classification mathematical models for pneumatic data into the wind tunnel operation system to perform real-time and automatic classification judgment on the pneumatic data generated by the wind tunnel test. Specifically, it is used for: Construct a binary classification mathematical model for judging whether pneumatic data is normal or abnormal by using the Isolation Forest method. Select normal pneumatic data and abnormal pneumatic data as training samples, and ensure that the proportion of abnormal data does not exceed 20%. Mix the normal and abnormal data disorderly, and do not use the data labels of normal or abnormal data during the training process; Construct a binary classification mathematical model for judging whether pneumatic data is normal or abnormal by using the One-Class Support Vector Machine. Only select normal pneumatic data as training samples; Construct a binary classification mathematical model for judging whether pneumatic data is normal or abnormal by using a deep neural network. Select normal pneumatic data and abnormal pneumatic data as training samples, and use the data labels of normal or abnormal data during the training process; A real-time judgment module, configured to, when all the multiple binary classification mathematical models for pneumatic data judge that the pneumatic data generated in real time by the wind tunnel is normal, determine that the pneumatic data generated by this wind tunnel test is normal and proceed to the next state of the wind tunnel test; otherwise, determine that there may be an abnormality in this wind tunnel test, remind the wind tunnel staff to check the problem, and make a decision manually.

3. An auxiliary decision-making device for wind tunnel operation based on pneumatic data monitoring, characterized in that, It includes: A memory, a processor, and a computer program stored on the memory and executable on the processor. When the computer program is executed by the processor, it implements the steps of the wind tunnel operation auxiliary decision-making method based on pneumatic data monitoring as described in claim 1.

4. A computer-readable storage medium, characterized in that, An information transmission implementation program is stored on the computer-readable storage medium. When the program is executed by the processor, it implements the steps of the wind tunnel operation auxiliary decision-making method based on pneumatic data monitoring as described in claim 1.

Images